DEF CON 32 - Exploiting Voice Cloning in Adversarial Simulation - Mark Foudy

We live in an era where voice verification is increasingly adopted in security protocols. The potential for abuse through voice cloning technology presents a significant and growing threat to cybersecurity. This talk dives into the alarming capabilities of deep learning to create highly convincing voice clones. Using my own voice as a case study, I will demonstrate a recorded simulation where the cloned voice successfully bypasses several major institutions’ voice verification systems. This presentation will outline the tools and techniques leveraged for voice cloning, discuss the pressing risks involved, and explore strategic countermeasures for red teams. The aim is to equip offensive security researchers with a nuanced understanding of voice cloning technology, emphasizing its implications for threat emulation and defensive strategy formulation. Attendees will gain insight into adversary tactics using publicly available voice samples for simulating voice-based attacks, providing a clear perspective on preparing defenses against such AI-driven threats.